Comparator propagation delay

[hithesh123]

How does a comparator prop delay relate to its speed(Bandwidth).
I googled but could not find any useful material.

Prop delay is just the delay from input to output. It doesn't say anything about max frequency input or Bandwidth of the comparator.

If you think of the comparator as an op amp with high gain, the gain should roll off at some frequency. I can't find any gain Vs. freq plots in the datasheets.

 

[FvM]

I agree, that tpd doesn't tell about maximum toggle rate or comaparator "bandwidth". But practically, both specification are be related somehow. You'll find a maximum toggle rate or minimum pulse width specification only for some fast comparators, e.g. ADCMP551. From these data and general considerations, an input signal should have at least a duration similar to tpd to cause the output to change it's state. A signal level specification (mostly called "overdrive") is also needed with tpd.

I assume, in most applications, a tpd value consderably below the signal cycle time will be required.

 

[dick_freebird]

Your 1/Tpd is a minimum bandwidth (so long as you have
application conditions similar to test conditions). There is
some degree of "pipelining" so you can probably pass a
signal that exceeds this frequency. However since Tpd
is 50%-50% you could see less than full scale output
swing (never fully settled).

Googling is the beginning, not the end, of research. A
lot more information resides at the manufacturers of the
devices. Application notes are meant to educate.

 

[hithesh123]

dick_freebird,

Thanks for the reply.
By googling, I didn't mean just looking at websites. I did read app notes.
I haven't come across an app note that clearly relates or even talks about bandwidth and prop delay.


FvM,

Tpd should be considerably below signal cycle time. But by how much?

 

[FvM]

Tpd should be considerably below signal cycle time. But by how much?

I've saying this without referring to a possible relation to comparator toggle rates. Just meaning, that the application mostly needs either a fast comparator reaction, or even more important, a reaction with predictable delay. The latter point addresses delay skew and signal dependant delay.

For the maximum toogle rate, respectively cycle time, the said device data suggests tcyc >= 2 tpd.

 

[leo_o2]

The coparator delay do not always depent on gain bandwidth.
More possibly, it depends on slew rate and large signal behavior.

How does a comparator prop delay relate to its speed(Bandwidth).
I googled but could not find any useful material.

Prop delay is just the delay from input to output. It doesn't say anything about max frequency input or Bandwidth of the comparator.

If you think of the comparator as an op amp with high gain, the gain should roll off at some frequency. I can't find any gain Vs. freq plots in the datasheets.

 

[varunkant2k]

Hi All, Sorry it may be too late to reply, but I felt the thread was still incomplete.
It will be beneficial if I elaborate more to make meaningful thread.

So from opamp (comparator) ft = UGB = f3-db*Adc ( where f3-db: 3dB bandwidth, Adc = DC gain of OPAMP)
Time constant of comparator = 1/(2*PI*f3-db). If we assume Comparator as a first order RC delay, it will take around 6 time constant to settle to final value.
Thats why Prop. Delay = 6* Time constant = 1/f3-db = UGB/Adc.

 

[hithesh123]

Hi All, Sorry it may be too late to reply, but I felt the thread was still incomplete.
It will be beneficial if I elaborate more to make meaningful thread.

So from opamp (comparator) ft = UGB = f3-db*Adc ( where f3-db: 3dB bandwidth, Adc = DC gain of OPAMP)
Time constant of comparator = 1/(2*PI*f3-db). If we assume Comparator as a first order RC delay, it will take around 6 time constant to settle to final value.
Thats why Prop. Delay = 6* Time constant = 1/f3-db = UGB/Adc.

That actually makes sense, at least for one comparator - MAX9144, 40ns prop delay.
so 6*40ns = 240ns.
f3db = 1/240 = 4.1MHz
which is what I got on breadboard.

 

[FvM]

I wonder how you define comparator 3 dB bandwidth and how you measure it in practice?

The expression refers to gain-bandwidth-product if I understand right

ft = UGB = f3-db*Adc

In case of MAX9144, GBW at 4 MHz signal frequency is > 200 MHz, because 100 mVpp input results in 5 Vpp output.

 

[varunkant2k]

For single pole system, f3-db refers to first pole frequency.
For 60dB DC gain opamp, with 200MHz UGB, first pole is at 200MHz/1000 = 200KHz.
That gives Prop. Delay = 1/200KHz = 5us.

 

[FvM]

For single pole system, f3-db refers to first pole frequency.
For 60dB DC gain opamp, with 200MHz UGB, first pole is at 200MHz/1000 = 200KHz.
That gives Prop. Delay = 1/200KHz = 5us.

Thanks for explaining your calculation. Obviously the result doesn't apply to MAX9144 which has 40 ns propagation delay, as reported below. I guess, you'll observe a similar mismatch for other standard comparators.

I basically agree to leo_o2, that there's no direct relation between small signal bandwidth and propagatiopn delay. That's particularly true for comparators that are basically uncompensated OPs thus the assumption about single pole frequency characteristic is void. Although it applies for unity gain compensated OPs, the calculation may be still dwarted by large signal effects.

 

[varunkant2k]

I basically agree to leo_o2, that there's no direct relation between small signal bandwidth and propagatiopn delay.

Hi FVM, Good Morning,
Here I gave explanation for any of device, whether compensated or not. Their 3-db bandwidth is always related to time constant of device.
One confusion usually happens for the term "Bandwidth" in compensated opamp and in comparator.
Generally Comparators are uncompensated and defining Unity-Gain_bandwidth for comparator is no use. For comparator Bandwidth refers to 3-dB (mostly).
For any systems 3-dB is directly related as UGB/Adc. So if device is compensated OPAMP, we know the UGB, and infact 3-db frequency.

 

[varunkant2k]

I am attaching one pdf from book -Analog Integrated Circuit Design by Johns and Martin.
It is explaining the Propagation delay of Comparators.

 

[FvM]

Yes, I see. The diagram and calculation illustrates well what's said before: The assumed single pole frequency characteristic doesn't apply to (uncompensated) comparators, large signal effects aren't considered.

 

[soner86]

Hi,
I have exactly the same problem for not an opamp but a voltage translator but the logic should be the same.
There is no freq. vs gain curve, no rise time. What I only have is the propogation delay and I need BW.

@varunkant2k,

Time constant of comparator = 1/(2*PI*f3-db). If we assume Comparator as a first order RC delay, it will take around 6 time constant to settle to final value.

I agree as f3-db≈1/6RC.
However, I do not agree with the following:

Thats why Prop. Delay = 6* Time constant = 1/f3-db.

It is not about propogation delay, it is about rise/fall time.
If you derive the RC circuit, and solve it for %10-%90 rise time, you get RC=RiseTime/2.2 and as we know that
f3-db≈1/6RC; we get RiseTime = 0.35/f3-db.

I couldn't find any relationship between propogation delay and f3-db anywhere. Can you please help me?